Distributed energy supply systems are most efficiently designed by mathematical optimization. However, optimization models often assume availability of all components at any time. In practice, security of energy supply is crucial; thus, reliability is mandatory but often neglected in optimization and only implemented subsequently employing expensive rules of thumb. In this work, we propose an exact optimization approach to identify (n - 1)-reliable designs for energy systems. The approach guarantees energy supply during the failure of 1 component at any time and is independent of probabilities and the selection of scenarios. (n - 1)-reliability is also necessary to allow for maintenance of components. For problems with high computational effort, we propose the inexact but computationally efficient (n - 1 max)-reliability approach which also guarantees energy supply but allows overproduction. A real-world case study shows that both approaches identify reliable designs at only a small increase of the total annualized costs compared to the unreliable base case. (C) 2018 Elsevier Ltd. All rights reserved.